Exploring Dynamic Scenes: 3D CNN-Based Activity Recognition
Shreshth Sharma, Misbah Anjum · 2024
The necessity of precisely extracting 3D models from freehand sketches is discussed in this study, especially in light of the expanding number of these models that are available online. It presents a research strategy that integrates convolutional neural networks (CNNs) with interactive attention to enhance the accuracy of 3D model retrieval by taking form distribution characteristics and semantic aspects into account. It also presents a novel method for 3D CNN-based activity retrieval in films, highlighting the importance of 3D CNNs in activity retrieval for video analysis and human activity recognition. Through empirical data, the research highlights the efficacy of the 3D CNN-based activity recognition model and investigates the architecture's capacity to capture dynamic movements and interactions inside video sequences. All things considered, this work offers fresh perspectives and approaches to the problems of 3D CNN-based activity recognition in video data.